AI Evaluators: Assessing a Shopping Assistant

What you'll do

You read interaction traces between real users and a shopping assistant on the project's own platform. Your job is to find where the assistant goes wrong: factual slips, broken logic, or product recommendations that do not help the person asking. From those findings you write structured rubrics and verifiers, meaning repeatable criteria that can judge response quality consistently later. The ad says the output feeds directly into improving the underlying model.

Building the evaluation framework is a big part of the role, not just marking answers. The ad expects you to create it from scratch, so you will be deciding what "good" looks like for e-commerce queries.

Who fits

• Quality assurance specialists, AI data evaluators and e-commerce professionals with an eye for detail
• Prior experience in prompt engineering, complex data annotation or software testing is welcome
• Comfort analysing text conversations in depth
• The ad does not state a degree requirement, a country restriction or a working language (the traces are text, so strong written English is a reasonable assumption)

What it pays

USD 30 per hour, as listed by Terac. The ad describes an ongoing engagement of 20+ hours per week, which would make it a steady role if the hours are available, but it does not guarantee them.

Worth knowing

Good:

• Ongoing work rather than a single session
• Rubric and verifier writing is a transferable evaluation skill
• Open worldwide, fully remote

Less good:

• 20+ hours a week is a real commitment and may not fit alongside another job
• USD 30 per hour is modest next to Terac's specialist studies
• The ad does not say how long the project runs

About this listing

Terac posted this on 2 September 2026 and lists it until 1 November 2026. The Apply button opens this study on Terac; you sign in or create a Terac account there to apply. The comments on the work are ours. See Terac.